[mpich-discuss] illegal ioctl

[Hans-Joachim Werner]

Thanks. Yes, most likely it runs out of some system resources. It happens independently
of the other memory the job needs. We use Global Arrays with MPI, and the problem happens
in ga_put (one-sided write to a global array). We can also run it with pure MPI, and the 
same problem occurs (it’s not SHMMAX). The application is the Molpro quantum chemistry program.
Joachim Werner
---
Prof. Dr. Hans-Joachim Werner
Institut for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart
Tel: +49 711 / 685 64400
Fax: +49 711 / 685 64442
email: werner at theochem.uni-stuttgart.de



Am 20.04.2014 um 16:52 schrieb Lu, Huiwei <huiweilu at mcs.anl.gov>:

> It’s possible your system is running out of memory or other system resources that cause the MPI job to abort.
> 
> —
> Huiwei Lu
> Postdoc Appointee
> Mathematics and Computer Science Division, Argonne National Laboratory
> http://www.mcs.anl.gov/~huiweilu/
> 
> On Apr 20, 2014, at 9:49 AM, Balaji, Pavan <balaji at anl.gov> wrote:
> 
>> 
>> This is not an MPICH problem.  It means that your application is asking MPICH to kill it using “MPI_Abort”.
>> 
>> — Pavan
>> 
>> On Apr 20, 2014, at 7:48 AM, Hans-Joachim Werner <werner at theochem.uni-stuttgart.de> wrote:
>> 
>>> On my system large jobs die with messages like this:
>>> 
>>> Last System Error Message from Task 3:: Inappropriate ioctl for device
>>> application called MPI_Abort(comm=0x84000004, 11) - process 3
>>> 
>>> Does anybody know how this can be avoided? Maybe a stack overflow?
>>> 
>>> Joachim Werner
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> discuss mailing list     discuss at mpich.org
>>> To manage subscription options or unsubscribe:
>>> https://lists.mpich.org/mailman/listinfo/discuss
>> 
>> _______________________________________________
>> discuss mailing list     discuss at mpich.org
>> To manage subscription options or unsubscribe:
>> https://lists.mpich.org/mailman/listinfo/discuss
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss